In this project we have built assets and techniques based on feedback control theory to help automate QoS management in software systems. Manual and ad-hoc methods of QoS management can be labor-intensive and error-prone. We apply proven, sound and robust techniques and methodologies from the mature field of control theory to design feedback control algorithms for managing the performance delivered by enterprise-class software. Control theory is especially appropriate for modern, on-demand IT systems because it excels at modeling and managing system dynamics.
Technology: robust and effective feedback control algorithms based on a sound mathematical theory
Components: AutoTune toolkit (in Java) provides a pluggable architecture for managing a wide range of systems. We have also developed embedded components (in C) that have been used for DB2.
Methodology & education materials
- A data-drive methodology for designing feedback control loops for software IT systems
- First-ever textbook on control theory targeted for computer science students and software engineers
- Classes taught at Columbia and UC Berkeley
Success stories
- Lotus Domino, Apache
- DB2: Utilities Throttling (v8.1), Self-tuning Memory (v9)
Last updated 18 Jul 2006